DeepSeek V4 Flash Vision Exp
DeepSeek V4 Flash Vision Exp is a large language model made by DeepSeek, added to the catalog on 21 Aug 2026. It has a context window of 1,048,576 tokens and costs $0.216 per 1M input tokens and $0.647 per 1M output tokens. It accepts text and images and produces text. Typical uses: answering questions about images, working with long text inputs and reasoning over text and visual context.
Price checked ยท source: OpenRouter
What DeepSeek V4 Flash Vision Exp is
DeepSeek V4 Flash Vision Exp is an experimental vision-enabled model from DeepSeek. It accepts text and images and produces text. Its image understanding extends the text capabilities of DeepSeek V4 Flash 0731, according to the provider. It can be used for text-only requests or prompts that pair written instructions with visual material, such as asking questions about an image or including an image as context for a text task.
The model supports reasoning, tools, structured outputs, and response formats. These capabilities suit tasks that involve working through a prompt, returning information in a specified structure, or using tools within a connected workflow. The listed capabilities do not specify particular tools, output schemas, or image formats, so those details depend on the surrounding application and prompt.
Its context window is 1,048,576 tokens, and its maximum output is 262,144 tokens. The context capacity can accommodate substantial text input alongside a task, while image input allows visual material to be part of the request. This combination is relevant to long-document work, image-based question answering, and tasks that need a structured text response. The experimental designation is part of the model's name and description; the catalog does not provide further details about its experimental status.
Key features
- Accepts text and image inputs and produces text
- Has a context window of 1,048,576 tokens
- Supports outputs of up to 262,144 tokens
- Supports reasoning, tools, and structured outputs
- Supports response formats
Typical uses
- Answering questions about images
- Working with long text inputs
- Reasoning over text and visual context
- Generating structured responses
- Tool-assisted workflows
DeepSeek V4 Flash Vision Exp pricing
US dollars, list price through OpenRouter.
| Item | Standard |
|---|---|
| Input, per 1M tokens | $0.216 |
| Output, per 1M tokens | $0.647 |
| Cached input read, per 1M tokens | $0.0069 |
The price last changed on 20 Nov 2026. One price change has been recorded since DeepSeek V4 Flash Vision Exp was added.
| Date | Input /1M | Output /1M | Change |
|---|---|---|---|
| $0.216 | $0.647 | Price lowered | |
| $0.44 | $1.32 | Added at this price |
| What a task costs | Cost |
|---|---|
| Answer a short question500 in / 300 out | $0.0003 |
| Summarise a 50-page PDF25,000 in / 800 out | $0.0059 |
| Review a 2,000-line source file30,000 in / 2,000 out | $0.0078 |
| Read a 300-page contract set150,000 in / 1,500 out | $0.033 |
DeepSeek V4 Flash Vision Exp specifications
| Model id | deepseek/deepseek-v4-flash-vision-exp |
|---|---|
| Version | deepseek-v4-flash-vision-exp-20260821 |
| Context window | 1,048,576 tokensAbout 786,432 English words |
| Max output | 262,144 tokens |
| Knowledge cutoff | Not published |
| Added to catalog | 21 Aug 2026 |
| Input types | text, image |
|---|---|
| Output types | text |
| Tool calling | Yes |
| Structured outputs | Yes |
| Reasoning effort control | Yes |
| Content moderation | Not applied by provider |
Sample outputs
The same prompts are given to every model, and the unedited output is shown with the date it was produced. Upend does not publish leaderboard scores it did not run.
No output has been recorded for DeepSeek V4 Flash Vision Exp yet.
Compare DeepSeek V4 Flash Vision Exp
Side by side with other models: prices, context window, what each accepts and supports, and the same four tasks priced on both.